I am working on a project where I have a number of Agents that are managed by a Team agent in coordination mode.
Each agent performs a specific task and I prefer to develop and test (eval) it separately. Each agent takes a number of inputs where some inputs come from other agents (I generally have them emit response_model for structured outputs). The issue I’m having is understanding the Agno idiomatic approach for having agents that receive inputs when they act stand-alone as well as in a team.
When the agent runs stand-alone alone, it’s easy to provide the data as session_state, via a message or as state injection in the prompt. However when it runs in a team you can’t set the input as a message or as session_state since the Agent is running as a team member and data is provided by other agents during runtime.
Having read the documents and particularly the one on Team state but it’s not really clear to me how the coordinating agent is supposed to make the output from an agent available to another agent and how the receiving agent would index/use this data?
Hi @anders-freeflow, thanks for reaching out and supporting Agno. I’ve shared this with the team, we’re working through all requests one by one and will get back to you soon.
If it’s urgent, please let us know. We appreciate your patience!
hey @anders-freeflow for making state shareable across team, we can use team_session_state it allows state to persist across the entire team Team State - Agno.
correct me if i am understanding it wrong , thanks
Hi @ayush, thanks for getting back to me. Two questions:
When working in a team, how do I get the coordinating team agent to store data in the tean session state automatically so that other team sub agents have access to it?
My original question is about: I am developing a stand-alone agent that can also work as part of a team, how can I give this agent the information it needs when it runs stand-alone or while in a team. Is there a single method/pattern for how to do this? I don’t assume that team session state works for a standalone agent?
The reason I ask is that I’d like to do evals on the standalone agent by providing it the data it needs the same way that the coordinating team agent will during run time. I can’t figure out what the common/shared method is for doing this for an agent that can run stand-alone as well as in a team.
